metadata
license: unknown
datasets:
- PolyAI/minds14
language:
- en
metrics:
- accuracy
- wer
- f1
- bleu
base_model:
- openai/whisper-tiny
pipeline_tag: automatic-speech-recognition
model-index:
- name: whisper-mind14-enUS
results:
- task:
type: ASR
dataset:
name: minds-14
type: enUS
metrics:
- name: Accuracy
type: Accuracy
value: 62.25
- task:
type: ASR
dataset:
name: minds-14
type: enUS
metrics:
- name: wer
type: wer
value: 0.38%
- task:
type: ASR
dataset:
name: minds-14
type: enUS
metrics:
- name: f1
type: f1
value: 0.6722
- task:
type: ASR
dataset:
name: minds-14
type: enUS
metrics:
- name: bleu
type: bleu
value: 0.0235
this model based on whisper-tiny model that trained with minds-14 dataset, only trained in english version : enUS
example of using model to classify intent:
>>> from transformers import pipeline
model_id = "kairaamilanii/whisper-mind14-enUS"
transcriber = pipeline(
"automatic-speech-recognition",
model=model_id,
chunk_length_s=30,
device="cuda:0" if torch.cuda.is_available() else "cpu",
)
audio_file = "/content/602b9a90963e11ccd901cbd0.wav" # Replace with your audio file path
text = transcriber(audio_file)
text
example output:
{'text': "hello i was looking at my recent transactions and i saw that there's a payment that i didn't make will you be able to stop this thank you"}